Most security teams already track dozens of KPIs, yet still struggle to say what they will change this quarter and how anyone will know it worked. Objectives and Key Results (OKRs) are a goal-setting method that answers that question: a short list of ambitious, qualitative objectives, each with a few measurable key results graded at the end of a cycle. Used well, they turn a security strategy and risk appetite into quarterly commitments; used badly, they become a renamed task list. This guide covers how OKRs differ from KPIs and KRIs, how to derive security objectives from strategy and the NIST CSF 2.0 Target Profile, how to write key results that measure outcomes, worked examples by domain, scoring, cadence and governance, and the common traps. It is written for CISOs, security programme managers and team leads who set goals; for building the measures themselves, see the Security Metrics and KPIs Programme Guide.
Key takeaways
- OKRs set direction, KPIs monitor health. A KPI tells you whether a control is performing; an OKR says which few things must change this cycle, and often borrows a KPI as its key result.
- Derive objectives from strategy and risk appetite. Start from the gaps between your NIST CSF 2.0 Current and Target Profiles and the risks outside appetite, not from the tools you want to buy.
- Key results measure outcomes, not activity. "Deploy phishing-resistant MFA" is a task; "95% of privileged accounts sign in only with phishing-resistant MFA" is a key result.
- Separate committed from aspirational. Regulatory deadlines and audit findings are committed OKRs expected to score 1.0; stretch goals are aspirational and a score around 0.7 is healthy.
- Keep OKRs out of performance ratings. Google's own guidance says OKRs are not employee evaluations; tying them to pay encourages sandbagging and gaming of security numbers.
What OKRs are, and how they differ from KPIs and KRIs
Google's re:Work guide describes objectives as ambitious goals that "may feel somewhat uncomfortable", and key results as measurable milestones that "should be easy to grade with a number". It recommends three to five objectives, with around three key results each, graded on a 0.0 to 1.0 scale. The OKR Playbook published by What Matters distinguishes two kinds: committed OKRs, which the team agrees will be achieved and are expected to score 1.0, and aspirational OKRs, with an expected average score of 0.7 and high variance.
| Instrument | Question it answers | Time frame | Security example | Who owns it |
|---|---|---|---|---|
| OKR | What must change this cycle, and how will we know? | Quarter or year | Objective: make account takeover of administrators impractical. KR: 100% of Tier 0 admin accounts use phishing-resistant MFA | A team or programme lead |
| KPI | Is a control or process performing as expected? | Continuous, reported monthly or quarterly | Median days to remediate KEV-listed vulnerabilities on internet-facing assets | Control owner |
| KRI | Is exposure inside risk tolerance? | Continuous, with thresholds | Number of internet-facing assets with a KEV vulnerability older than its due date | Risk owner |
| Project milestone | Was a deliverable completed? | Project plan | Hardware security keys issued to the infrastructure team | Project manager |
The practical rule: when a KPI or KRI is outside its target or tolerance and fixing it needs a change in how people work, that is an OKR candidate. When it is inside tolerance, it stays a KPI on the dashboard. The metrics, data sources and definitions behind each number belong in the measurement programme, which NIST SP 800-55 Volume 1 (final, December 2024) covers.
Deriving security objectives from strategy
An OKR set should trace upward to something leadership already agreed. Three inputs do most of the work:
- Strategy: the UK Cyber Governance Code of Practice (published 8 April 2025) asks boards to gain assurance that a cyber strategy is "aligned with, and embedded within" the wider organisational strategy (action B1) and that it is being delivered and achieving its intended outcomes (B4). Annual OKRs are a practical way to show B4. See the Security Strategy and Roadmap Guide.
- Risk appetite: the same Code asks boards to define the cyber risk appetite (A3) and agree tolerances for metrics (E2). Any KRI outside tolerance is a natural objective. See the Risk Appetite and Tolerance for Cyber Guide.
- CSF 2.0 profiles: NIST CSWP 29 defines a Target Profile as the desired outcomes an organisation "has selected and prioritized" and tells organisations to run a gap analysis between Current and Target Profiles, then record the prioritised plan in a risk register or POA&M. The highest priority gaps become annual objectives; the actions for this quarter become key results. See the NIST CSF 2.0 Self-Assessment.
- Baseline practices for smaller teams: CISA's Cross-Sector Cybersecurity Performance Goals are a voluntary baseline organised by the six CSF functions and designed to help organisations prioritise a limited number of essential actions. Unmet goals make good first-year objectives for teams without a full CSF profile.
- List candidate outcomes: top CSF gaps, KRIs out of tolerance, open audit or regulator findings, incident lessons and business change (new markets, acquisitions, AI adoption).
- Rank by risk reduction and business dependency: keep three to five annual objectives for the function, phrased as outcomes a non-specialist executive understands.
- Classify each: committed (deadline driven, must be met) or aspirational (stretch).
- Cascade carefully: team OKRs should contribute to a function objective, but not every team needs one per objective. Forced cascades produce artificial key results.
- Check capacity: CSF 2.0 GV.RR-03 expects resources commensurate with the strategy. If an objective has no funded people behind it, either fund it or drop it; see the Security Budget and Investment Planning Guide.
Writing key results that measure security outcomes
The OKR Playbook says key results must describe outcomes, not activities, and that each must be possible to grade objectively at the end of the quarter. In security, the commonest failure is a key result that measures effort ("run four phishing simulations") rather than the change in exposure or capability.
| Weak key result (activity) | Why it is weak | Stronger key result (outcome) |
|---|---|---|
| Roll out the new PAM tool | Deployment says nothing about risk; a tool can be installed and unused | 90% of domain admin sessions brokered through PAM with recorded sessions, measured weekly |
| Run four phishing simulations | Counts activity; the number of tests is in the team's own control | Report rate on simulations rises from 35% to 60% and median time to first report falls below 10 minutes |
| Review all service accounts | "Review" has no measurable end state | 100% of production service accounts have a named owner and 80% have credentials rotated in the last 90 days |
| Improve vulnerability management | Not measurable at all | Zero KEV-listed vulnerabilities on internet-facing assets past their due date at quarter end |
| Write an incident response plan | A document is an output, not a capability | Two tabletop exercises completed with executives, and 100% of actions from the last exercise closed |
- State the baseline: "from 62% to 90%" is gradable; "to 90%" invites an argument about where you started.
- Name the population: say which assets, accounts or applications are in scope, and report coverage beside the result so the denominator cannot shrink quietly.
- Pair speed with quality: if one key result pushes for faster triage, add a second that guards false negative rates or escalation accuracy, so the team cannot win one by losing the other.
- Prefer leading indicators for a quarter: incident counts are noisy over 90 days; coverage, time to remediate and control effectiveness tests move faster and are within the team's influence.
- Avoid targets the team does not influence: "zero breaches" is not a key result; it is a hope.
Worked security OKR examples by domain
The examples below use realistic structures. Replace the percentages with your own baselines; the numbers here are illustrative, not benchmarks.
| Domain | Objective | Key results (illustrative) | Type |
|---|---|---|---|
| Workforce identity | Make administrator account takeover impractical | KR1: 100% of Tier 0 admins use phishing-resistant MFA. KR2: standing global admin roles reduced from 18 to 4. KR3: 100% of admin role activations are time-bound and approved | Committed |
| Non-human identity | Know and control every production machine credential | KR1: 95% of service accounts, API keys and tokens in the inventory have an owner. KR2: median age of production secrets falls from 400 to 90 days. KR3: secrets found in code repositories are revoked within 24 hours in 90% of cases | Aspirational |
| Vulnerability management | Close the window attackers use on the internet edge | KR1: zero KEV-listed vulnerabilities past due on internet-facing assets. KR2: 98% of internet-facing assets covered by authenticated scanning. KR3: median time to patch critical edge devices under 7 days | Committed |
| Detection and response | Find intrusions ourselves, before a third party tells us | KR1: 100% of identity provider, EDR and cloud control plane logs feed the SIEM. KR2: detection rules mapped to the top 20 ATT&CK techniques for our threat model, each tested once. KR3: mean time to contain a confirmed endpoint compromise under 4 hours | Aspirational |
| Resilience | Prove we can recover critical services without paying a ransom | KR1: isolated, immutable backups for 100% of tier 1 systems. KR2: two full restore tests meeting recovery time objectives. KR3: executive ransomware tabletop held with decisions recorded | Committed |
| Third parties | Reduce exposure from suppliers with access to our data | KR1: 100% of critical suppliers assessed this cycle. KR2: vendor accounts without activity in 90 days disabled. KR3: contract security clauses present for 90% of critical suppliers | Committed |
| Security culture | Make reporting suspicious activity the norm | KR1: phishing simulation report rate from 35% to 60%. KR2: median report time under 10 minutes. KR3: 80% of staff rate the reporting process as easy in a survey | Aspirational |
For deeper measures behind these key results, see the Identity Security Metrics Guide, Vulnerability Management Metrics and Security Operations Metrics Guide.
Scoring, cadence and governance
Google grades key results individually and roughly averages them to grade the objective. Its re:Work guide puts the sweet spot for ambitious OKRs at 60 to 70%: teams that consistently score higher are not setting stretching goals. That logic only applies to aspirational OKRs; a committed security OKR, such as meeting a regulatory deadline, that scores below 1.0 needs an explanation and a recovery plan.
| Cadence point | What happens | Security specific note |
|---|---|---|
| Annual planning | Set 3 to 5 function objectives from strategy, risk appetite and CSF gaps | Align with the budget cycle and the board's annual risk review |
| Quarter start | Set key results with baselines, owners and data sources | Confirm the data can be produced before committing; a KR you cannot measure cannot be graded |
| Fortnightly or monthly check-in | Confidence rating per KR (on track, at risk, off track) | Re-plan when an incident consumes capacity rather than silently letting OKRs slip |
| Quarter end | Grade 0.0 to 1.0, write a short retrospective, carry forward or drop | Committed OKRs below 1.0 go to the risk register with an owner and date |
| Board or executive report | Show objective progress next to KRIs and tolerances | The Cyber Governance Code asks for formal reporting at least quarterly (E2); see the Board Cyber Risk Reporting Guide |
- Transparency: Google makes OKRs public inside the organisation. For security, publish objectives and progress widely, but keep detail that maps your weaknesses (for example, which critical systems lack backups) to a restricted audience.
- Separation from pay: re:Work states that OKRs are not synonymous with employee evaluations. In security this matters doubly: linking bonuses to "fewer incidents" discourages reporting.
- Interrupt work: security teams lose capacity to incidents and urgent advisories. Reserve capacity explicitly, or record an interrupted quarter honestly rather than rewriting key results after the fact.
How to introduce security OKRs: a phased rollout
- Pilot one quarter with one team: choose a team with good data, such as vulnerability management or identity. Write two objectives, no more than three key results each.
- Baseline every key result before the quarter starts: if the baseline takes more than a week to produce, the key result is not ready.
- Run check-ins and a retrospective: record what was graded, what the score meant and which key results drove the wrong behaviour.
- Extend to the function: set three to five function objectives for the year; let teams propose key results that contribute.
- Connect to governance: report OKR progress beside KPIs and KRIs to the security steering group, then quarterly to the executive or board.
- Roll back if needed: if OKRs are duplicating the project portfolio or the KPI pack, stop writing new ones for a quarter, keep the KPIs, and restart with fewer, outcome-only objectives.
Small versus large organisations: a team of five should run one set of function OKRs with two or three objectives, reviewed monthly by the security lead and an executive sponsor. A large enterprise needs function OKRs, team OKRs and a light alignment review so that, for example, the identity team's objective and the cloud team's objective do not both claim the same key result.
Common failure patterns
- Business as usual dressed up: the OKR Playbook warns against OKRs based on what the team can already do without changing its work. "Keep patching" is a KPI, not an objective.
- "Who cares?" objectives: an objective whose perfect score brings no clear business benefit, such as "migrate SIEM to new platform", should be rewritten around the benefit, for example faster detection of identity attacks.
- Key results that miss the objective: the Playbook notes that a team can score 1.0 on every key result and still miss the intent. Test this: if all key results are met, will a reasonable executive agree the objective is achieved?
- Shrinking denominators: MFA coverage improves because accounts were removed from scope, not because MFA was deployed. Always report the population counted.
- Too many OKRs: ten objectives with five key results each is a project list; re:Work recommends three to five objectives.
- Goals linked to bonuses: once pay depends on the number, teams set easy targets or redefine incidents, and the data stops being trustworthy.
Measuring whether OKRs are working
- Outcome share: percentage of key results that measure an outcome rather than an activity, reviewed at planning (target: over 80%).
- Traceability: percentage of objectives that trace to a CSF Target Profile gap, a KRI outside tolerance or a strategy commitment.
- Score distribution: committed OKRs scoring 1.0, and aspirational OKRs averaging near 0.6 to 0.7 over several quarters.
- KRI movement: number of KRIs brought back inside tolerance by OKR work over the year.
- Re-plan honesty: number of key results changed mid-quarter without a recorded reason (target: zero).
Framework mapping
| Framework | Reference | How OKRs support it |
|---|---|---|
| NIST CSF 2.0 | GV.OC-01; GV.RM-01; GV.RM-02 | Objectives informed by mission, agreed risk management objectives and risk appetite and tolerance |
| NIST CSF 2.0 | GV.OV-01; GV.OV-02; GV.OV-03 | Quarterly grading reviews strategy outcomes and performance and adjusts direction |
| NIST CSF 2.0 | GV.RR-03; ID.IM-01 | Resourcing commensurate with strategy; improvements identified from evaluations feed the next cycle |
| NIST SP 800-53 Rev 5 | PM-6 Measures of Performance; PM-9 Risk Management Strategy; PM-31 Continuous Monitoring Strategy | Key results draw on programme measures and monitoring; objectives implement the risk strategy |
| ISO/IEC 27001:2022 | Clause 6.2 Information security objectives; Clause 9.1 Monitoring, measurement, analysis and evaluation | OKRs are one way to set measurable objectives with plans, owners, timing and evaluation |
| UK Cyber Governance Code of Practice | A3, B1, B4, E2 | Risk appetite, strategy alignment, evidence that the strategy delivers, quarterly reporting with metrics and tolerances |
Practitioner checklist
- Map each candidate objective to a strategy commitment, CSF Target Profile gap or KRI outside tolerance.
- Limit the function to three to five objectives with around three key results each.
- Label every OKR committed or aspirational and agree what score is expected.
- Write key results as outcomes with a baseline, a target, a population and a data source.
- Pair speed targets with quality guards so key results cannot be gamed.
- Hold check-ins at least monthly and record any mid-quarter change with a reason.
- Grade at quarter end, move missed committed OKRs to the risk register and write a short retrospective.
- Report OKR progress next to KPIs and KRIs to the executive at least quarterly.
- Keep OKRs out of individual performance ratings and bonuses.
Standards and references
- NIST: The NIST Cybersecurity Framework (CSF) 2.0, CSWP 29
- NIST: SP 800-55 Vol. 1, Measurement Guide for Information Security
- GOV.UK: Cyber Governance Code of Practice
- CISA: Cross-Sector Cybersecurity Performance Goals
- Google re:Work: Set goals with OKRs
- What Matters: Google's OKR Playbook
Related NHI Mgmt Group resources: Security Metrics and KPIs Programme Guide · CISO Guide: First 90 Days · Security Strategy and Roadmap Guide · Board Cyber Risk Reporting Guide